Algorithmic ''centrifuge'' - page 5
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Oh, I get it! People must have thought I was suggesting that I was synthesising indicators during the optimisation process. This is not the case.
I am talking about ready-made indicators. I was saying that there is a base of indicators, formulas and algorithms that can be searched inside the Signal during optimization.
It is not about synthesizing new indicators. And there wasn't.
...
601 million - number of unique indicator pairs on 32 bars, without their parameters. Maximum - number of 32x32 matrices excluding reflections and squared
if each parameter has 2 (at least - just scaling linear+distortion), with integer values from 1 to 9 inclusive, then additionally multiply by 10^4
You have completely misunderstood the idea. It's about using existing CB indicators, not synthesising new ones.
It's amazing that even if it was about synthesizing new indicators, HOW did you manage to calculate it???
The 601 million unique pairs is taken from the ceiling. It's not clear what you were calculating. It's not clear at all...
ZS. However, it happens to everyone... Just understand that the essence of any strategy is a signal to enter and exit. The signal consists of 3 parameters (usually). All three parameters are derived from some indicators. Indicators are already written and included in the program. They only need to be enumerated inside the signal and look for the best values for each combination.
At the end we will get the best combination of indicators for the signal with the best values. And that, - is the strategy.
It will not affect the performance of real trading not on historical data. It is guaranteed to fail. It is only a matter of time. In this case, you really have to pay for the time of the supercomputer, because the usual one cannot cope with it.
Each Indicator can be represented by ONE Parameter placed as part of the Market Entry/Exit Signal.
9 indicators create 9 CONFIGURATIONS of the entry/exit signal, with 3 indicators included in ONE signal.
Why 9 and not 27?
Because, the variation (Indicator_1, Indicator_2, Indicator_3) within the signal can be mixed -(Indicator_2, Indicator_3, Indicator_1) and the essence of the signal will not change.
not 9 or 27, but 84. I think you'll see why.
Peter, all you can achieve with your centrifuge is a smoother and more perfect tester grail on historical data, on which the optimization of indicator fitting was performed.
It will not affect the performance of real trading not on historical data. It is guaranteed to fail. It is only a matter of time. In this case, you really have to pay for the time of the supercomputer, because the usual one cannot cope with it.
Nikolai, the tester is all we have. ))
So why can't an ordinary computer do the job? The same search for values for the parameters included in the signal.
not 9 or 27, but 84. I think you'll understand why.
Honestly, I don't. I don't understand why.
Simpler: 3 indicators - 5 configurations - one signal.
Example:
THREE indicator configurations in a Signal: (1,2,3) or (2,3,1) or (3,2,1) or (1,3,2) or (3,1,2)... ONE AND ONE.
So, there's a lot less to go through than I thought earlier.
Honestly, no. I don't see why not.
Simpler: 3 indicators - 5 cofigurations - one signal.
Example:
THREE indicator configurations in a Signal: (1,2,3) or (2,3,1) or (3,2,1) or (1,3,2) or (3,1,2)... ONE AND ONE.
So, there's a lot less to go through than I thought earlier.
To clarify, if I understand correctly. I have 9 different indicators, and if there may be 3 different indicators in a signal.
Then the total number of combinations without repeats = 9*8*7/6!
i.e. total number of combinations divided by the 6 factorial. 6! This is the number of permutations of three indices, and is just the elimination of equal combinations (1,2,3) or (2,3,1) etc.
Maxim Kuznetsov has a good understanding and will correct you if needed.
Let me explain, if I have understood correctly. In case of 9 different indicators, and if there can be 3 different indicators in the signal.
Then the total number of combinations without repeats = 9*8*7/6!
i.e. total number of combinations divided by the 6 factorial. 6! This is the number of permutations of three indices, and is just the elimination of equal combinations (1,2,3) or (2,3,1) etc.
Maxim Kuznetsovis well versed and can correct you if necessary.
84 combinations of groups of TRY, for 9 indicators? Without "useless" combinations? Seems to be too much... I think there's an error in the calculation. But, never mind.
The overkill on the signal parameters isn't too great. We're not talking about millions.
Another question is how to find the best values for each combination of indicators in the signal? You need to understand the mechanism itself. Maybe there is a catch...
It's not parameters (or rather, not only parameters), but combinations of blocks, which are sub-strategies and from which the final strategy is compiled. You may not have read my link, the quote below. If you read it, write your opinion, it will be interesting.
Igor, why bother opening the terminal, it's more interesting to think without it))).
The blocks must have switches to be combined in different ways. And the position of these switches can also be optimised, each optimisation pass will correspond to some combination of blocks. Here everything is happening in a big gap from reality.
For the blocks to be combined in different ways, they must have on/off switches. Well, the position of these switches can also be optimised, each optimisation pass will correspond to some combination of blocks. This is all happening here in a big break from reality.
The idea is exactly that each optimization pass will correspond to a different combination of blocks. If we introduce for every block one parameter T - variation of its internal pairs, and for every block we form a model so that when trying not so big number of T we can judge about wide range of module's work, for example T = 1...10. Then by trying all combinations of modules and their variations T we can judge about stability of this combination and select it for further detailed investigation (with big number of variations of pairs)